Skip to main content

2.1 The Distributed Mind

Stage II: The Graph

"A robot is not a singular entity. It is a choir."

1. The Distributed State Problem​

In Stage I, we learned about the Process (Chapter 1.1). Now, imagine we have 50 processes.

  • Process A reads the Camera.
  • Process B controls the Wheels.
  • Process C maps the room.

How does Process A tell Process C that it saw a wall? If Process A calls ProcessC.add_wall(), they are Coupled. If Process C crashes, Process A crashes. This is a Monolith. Monoliths die.

The Solution: The Graph​

We treat every process as a Node in a directed graph. Nodes do not know each other. They only know Topics (Edges).

$$ G = (V, E) $$

  • $V$: The set of Nodes (Computing Units)
  • $E$: The set of Topics (Data Streams)

2. Distributed Consistency (CAP Theorem)​

In a distributed system, you can only have 2 of the 3:

  1. Consistency: Everyone sees the same data at the same time.
  2. Availability: The system always responds to requests.
  3. Partition Tolerance: The system survives network cuts.

Robotics Choice: We usually sacrifice Consistency for Availability. If the LIDAR node sends data, the Mapping node might receive it 10ms late. That's fine. We tolerate Eventual Consistency. But if the robot drives into a wall because the Stop command didn't arrive, that's bad. This is why ROS 2 uses DDS (Data Distribution Service).

3. The Node Lifecycle​

A Node is not just a while True loop. It has a lifecycle.

  1. Unconfigured: Born, but empty.
  2. Inactive: Configured (params loaded), but asleep.
  3. Active: Processing data.
  4. Finalized: Dead.

Deep Concept: Managed Nodes. In ROS 2, a "System Manager" can tell a Driver Node: "Go to Sleep", and it stops publishing. This saves CPU/Battery without killing the process.

Deep FAQ​

Q: Why not use Shared Memory? A: We do! ROS 2 has "Zero Copy" transport (Iceoryx) for nodes on the same machine. But the abstraction remains Pub/Sub so that if we move the Node to a different computer over Wi-Fi, the code doesn't change.

Q: What is a Discovery Storm? A: When 100 nodes wake up at once, they all shout "I am here!" via Multicast. This can flood the network switch. This is why we use Discovery Servers in large fleets.